Index: Objects/unicodeobject.c
===================================================================
--- Objects/unicodeobject.c	(révision 85136)
+++ Objects/unicodeobject.c	(copie de travail)
@@ -1153,19 +1153,73 @@
     return ret;
 }
 
-static void
+/* Convert a Unicode object to a wide character string. Return the number of
+   wide characters including the nul character. w can be NULL. */
+static Py_ssize_t
 unicode_aswidechar(PyUnicodeObject *unicode,
                    wchar_t *w,
                    Py_ssize_t size)
 {
 #if Py_UNICODE_SIZE == SIZEOF_WCHAR_T
     memcpy(w, unicode->str, size * sizeof(wchar_t));
-#else
+    return size;
+#elif Py_UNICODE_SIZE == 2 && SIZEOF_WCHAR_T == 4
+    register const Py_UNICODE *u;
+    const Py_UNICODE *uend;
+    const wchar_t *worig, *wend;
+    Py_ssize_t nchar;
+
+    u = PyUnicode_AS_UNICODE(unicode);
+    uend = u + PyUnicode_GET_SIZE(unicode);
+    if (w != NULL) {
+        worig = w;
+        wend = w + size;
+        while (u != uend && w != wend) {
+            if (0xD800 <= u[0] && u[0] <= 0xDBFF
+                && 0xDC00 <= u[1] && u[1] <= 0xDFFF)
+            {
+                *w = (((u[0] & 0x3FF) << 10) | (u[1] & 0x3FF)) + 0x10000;
+                u += 2;
+            } else {
+                *w = *u;
+                u++;
+            }
+            w++;
+        }
+        if (w != wend) {
+            *w = L'\0';
+            w++;
+        }
+        return w - worig;
+    } else {
+        nchar = 1; /* nul character at the end */
+        while (u != uend) {
+            if (0xD800 <= u[0] && u[0] <= 0xDBFF
+                && 0xDC00 <= u[1] && u[1] <= 0xDFFF)
+                u += 2;
+            else
+                u++;
+            nchar++;
+        }
+    }
+    return nchar;
+#elif Py_UNICODE_SIZE == 4 && SIZEOF_WCHAR_T == 2
     register Py_UNICODE *u;
-    register Py_ssize_t i;
+    const wchar_t *worig, *wend;
+
+    if (w == NULL)
+        return size;
+
     u = PyUnicode_AS_UNICODE(unicode);
-    for (i = size; i > 0; i--)
+    worig = w;
+    wend = w + size;
+    while (w != wend) {
+        /* FIXME: create surrogate pairs if needed */
         *w++ = *u++;
+    }
+    return w - worig;
+#else
+#  error "unsupported wchar_t and Py_UNICODE sizes"
 #endif
 }
 
@@ -1178,17 +1232,7 @@
         PyErr_BadInternalCall();
         return -1;
     }
-
-    /* If possible, try to copy the 0-termination as well */
-    if (size > PyUnicode_GET_SIZE(unicode))
-        size = PyUnicode_GET_SIZE(unicode) + 1;
-
-    unicode_aswidechar(unicode, w, size);
-
-    if (size > PyUnicode_GET_SIZE(unicode))
-        return PyUnicode_GET_SIZE(unicode);
-    else
-        return size;
+    return unicode_aswidechar(unicode, w, size);
 }
 
 wchar_t*
@@ -1203,19 +1247,19 @@
         return NULL;
     }
 
-    if ((PY_SSIZE_T_MAX / sizeof(wchar_t) - 1) < PyUnicode_GET_SIZE(unicode)) {
+    buflen = unicode_aswidechar(unicode, NULL, 0);
+    if (PY_SSIZE_T_MAX / sizeof(wchar_t) < buflen) {
         PyErr_NoMemory();
         return NULL;
     }
 
-    buflen = PyUnicode_GET_SIZE(unicode) + 1; /* copy L'\0' */
     buffer = PyMem_MALLOC(buflen * sizeof(wchar_t));
     if (buffer == NULL) {
         PyErr_NoMemory();
         return NULL;
     }
-    unicode_aswidechar(unicode, buffer, buflen);
-    if (size)
+    buflen = unicode_aswidechar(unicode, buffer, buflen);
+    if (size != NULL)
         *size = buflen;
     return buffer;
 }